Method and device for distributing storage resources in GPU in big integer calculating process

Abstract

The invention discloses a method and device for distributing storage resources in a GPU in the big integer calculating process. The method includes the steps that in the calculating execution process of all sorts of big integers by the adoption of GPU threads, a register is distributed for each thread, in the big integer calculating executing process of the threads, the operating number and the middle calculating variable of big integer calculating executed currently by one thread are stored into the corresponding register distributed by the thread, and the modulus of big integer calculating executed currently by the thread and other data are stored into an internal memory. In this way, when the operating number and the middle calculating variable need to be called in the big integer calculating process in the GPU threads, the operating number and the middle calculating variable can be obtained by repeated calling of the distributed registers directly. On one hand, the register using quantity of the GPU threads is reduced, and the parallel execution efficiency of the threads is improved; on the other hand, necessary registers are reserved for big integer calculating, and the execution speed of big integer calculating can be increased more effectively.

Description

technical field [0001] The present invention relates to large integer calculation technology, in particular to a method and device for allocating storage resources when performing large integer calculation in a graphics processing unit (GPU, Graphics Processing Unit). Background technique [0002] The use of multi-core parallel computing is an important way to improve processor performance. Therefore, GPUs including massive parallel structure computing units have emerged. GPUs have developed into high-performance general-purpose processors with high parallelism, multi-threading, fast computing, and large memory bandwidth. The GPU architecture is divided into three layers: the first layer consists of several thread processor clusters (TPC, Thread Processing Cluster), the second layer consists of multiple stream multiprocessors (SM, Streaming Multiprocessor), the third The layer is a stream processor (SP, Stream Processor) that constitutes the SM, and can also be called a thre...

AI Technical Summary

Problems solved by technology

However, the data bit width of large integers in the public key cryptography algorithm is generally 128 to 2048 bits, and the calculation overhead of large integers is very large.

Images